Skip to content

v1.1.8 — Pydantic AI v2, Dependabot, CI hardening

Choose a tag to compare

View all tags
@ralfbecher ralfbecher released this 16 Jul 15:24
· 19 commits to main since this release
v1.1.8
dce8f9f
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

First stable release since v1.1.5. Upgrade directly from v1.1.5 to v1.1.8 — v1.1.6 and v1.1.7 were tagged but shipped broken and are superseded by this release (details below).

Highlights

  • Pydantic AI v2 migration. Moved from pydantic-ai 1.x to 2.10. OpenAIModel → OpenAIChatModel, and the MCP layer now uses MCPToolset over explicit StdioTransport / StreamableHttpTransport. MCP sampling.tools advertisement is preserved (pydantic-ai v2 still ships a bare SamplingCapability(), so it's re-enabled explicitly).
  • Dependency automation. Added Dependabot config covering the uv, GitHub Actions, and Docker ecosystems, plus the first batch of resulting updates (5 production dependency bumps, 8 GitHub Action major versions).
  • CI hardening. Test matrix across Python 3.11–3.13, and a new smoke step that boots the app and fetches a static asset — catching runtime-only "server starts but can't serve the UI" failures that unit tests miss.

Fixes folded in

  • MCP chat startup crash (regression from the v2 migration): the sampling wrapper referenced a pydantic-ai v1 attribute that MCPToolset doesn't have, so any configured MCP server raised AttributeError on chat start. Sampling errors now render correctly too (v2 returns ErrorData rather than raising).
  • Broken UI on Python 3.14: the 3.14 base image shipped an app whose static assets all returned HTTP 500 (anyio.NoEventLoopError, an upstream chainlit issue). Reverted to python:3.13-slim and capped requires-python to <3.14.

Docker

docker pull ralforion/orionbelt-chat:1.1.8   # or :latest

Runs on Python 3.13.

Full changelog: v1.1.5...v1.1.8

Assets 2
Loading